| config: configs/ar_l.yaml | |
| code_path: | |
| gpt_model: GPT-L | |
| gpt_ckpt: | |
| finetune: false | |
| model_weight_strict: true | |
| gpt_type: c2i | |
| ema: true | |
| cls_token_num: 1 | |
| dropout_p: 0.1 | |
| token_dropout_p: 0.1 | |
| drop_path_rate: 0.0 | |
| compile: true | |
| results_dir: | |
| data_path: ImageNet2012/train | |
| dataset: imagenet_code | |
| vq_path: | |
| vq_ema: false | |
| image_size: | |
| num_classes: 1000 | |
| epochs: 300 | |
| lr: 1e-4 | |
| warmup_steps: | |
| warmup_ratio: 0.05 | |
| weight_decay: 5e-2 | |
| beta1: 0.9 | |
| beta2: 0.95 | |
| max_grad_norm: 1.0 | |
| global_batch_size: 512 | |
| global_seed: 0 | |
| num_workers: 24 | |
| log_every: 50 | |
| ckpt_every: 5000 | |
| save_epochs: 2 | |
| mixed_precision: bf16 | |
| learnable_tok_embeddings: true | |
| resume_from_newest_ckpt: true | |
| debug_mode: false | |
| correction_prob: 0.3 | |
| rank: 0 | |
| world_size: 8 | |
| gpu: 0 | |
| dist_url: env:// | |
| distributed: true | |
| dist_backend: nccl | |